Software built with intent. Shipped with velocity.
A founder-led engineering studio in Durban. We find where your operation leaks time and money, then build the AI and software that fixes it, with the accuracy measured before you rely on it.
Measure first. Then build.
Three ways in, each with a fixed scope and a published price. Most clients start with the audit, so every rand spent on a build is aimed at a problem we have already measured.
We sit with your team, read your records and watch the real process. You get a written, costed answer on where AI will save or earn you money, and where it won't.
See what you get 02 · BUILDKnowledge assistants, document automation, WhatsApp AI and web platforms, each with accuracy tested on your own data before go-live.
Browse services 03 · RUNWeekly strategy time, one small build a month, and an independent second opinion before you sign the next software contract.
How it worksWhere two oceans converge.
Named after Cape Agulhas — the southernmost tip of Africa, where the Indian and Atlantic Oceans meet — our name marks the convergence of strategy, design, and engineering.
We partner with technical teams across Africa and beyond to design, build, and ship digital products of exceptional quality — engineered with precision, designed with intent, and built to scale.
“We don't sell AI hype — we sell measurable hours and rand saved. Every build is quoted in those units.”
Darrshan Govender
Builds production LLM systems — natural-language analytics, retrieval-augmented chatbots, multi-agent pipelines and test-time reasoning. The operating principle behind every build: treat AI like any other production system — structured outputs, eval-gated CI, guardrails by default, and full observability across cost and latency.
Production-grade, not demos.
InsightEngine
Natural-language-to-SQL analytics. Ask a question in plain English; get a governed query and an answer — with sqlglot AST guardrails so generated SQL can never run unsafe.
RAG Support Bot
Hybrid search + citation verification — every answer traces to a source. CI eval-gated for faithfulness, recall@k, latency and cost.
Churn & MRR Forecasting
XGBoost churn classifier + Prophet revenue forecasting — flagging at-risk accounts early and projecting MRR with confidence bands.
Thinking Loop
Five test-time reasoning strategies — chain-of-thought, self-consistency, decomposition, ToT-lite and direct — combined under an LLM adjudicator that picks the strongest answer.
Hybrid Recommender
Content + collaborative signals — more than double baseline recall (0.41 vs 0.18), even for cold-start users.
Semantic Cache
Recognises when two questions mean the same thing — cutting LLM token spend 30–80% on real chatbot workloads.
What we build.
Productised services with a fixed scope and a published price. Every one is held to the same rule: measured on your own data before anyone relies on it.
Knowledge assistants
A secure assistant over your own documents and archives. Every answer cites its source page, and it says so when the answer isn't there.
AIDocument automation
Invoices into Xero or Sage, proofs of delivery, contracts and quotes, read and checked automatically, with a person approving what matters.
AIWhatsApp AI concierge
An assistant on your business WhatsApp that answers routine questions in your voice and hands anything real to a person.
Operational audits
Independent audits of cash, payouts, stock and controls across multi-site operations, starting with fuel retail. Every finding in rand, traced to source.
Web platforms
Fast, credible websites and client portals that bring in enquiries, without a bloated CMS. Everything registered in your name.
AIAI retainer
Senior AI judgement on tap: weekly strategy time, one small build a month, and a second opinion before you sign the next contract.
From idea to launch.
A structured, transparent process that keeps you in the loop at every stage — no surprises, no long silences.
Discovery
We start with your business, your users and your goals. Workshops and research define the scope, technical requirements and the metrics we'll hold the work to.
Design
Wireframes, prototypes and polished UI that bring the vision to life. Every screen is reviewed with you before a single line of production code is written.
Develop
Production-grade code in two-week sprints, with live demos and a staging environment from week one. You see real progress every fortnight — not a big reveal at the end.
Launch & Support
We handle deployment, performance testing and go-live. After launch: monitoring, fixes and incremental features — and a clean handover whenever you want the keys.
Have something in mind? Let's make it real.
Start a projectThings people ask.
Where should we start?
Usually with the AI Operations Audit: two weeks, a fixed R15,000, and a written report on where AI will save or earn you money in your specific operation, and where it won't. It is designed to be useful even if you never build anything with us. If you already know exactly what you need, we can go straight to a build.
How long does a typical project take?
The audit takes two weeks. Most builds run three to ten weeks: a document pipeline can be live in a month, a knowledge assistant typically takes six to ten weeks. You get a concrete timeline before any work begins.
What does an engagement cost?
Every service has a published range. The AI Operations Audit is a fixed R15,000. Business websites start around R30,000, and most AI builds, such as knowledge assistants, document automation and WhatsApp assistants, fall between R25,000 and R120,000. Larger platforms are phased, with each phase priced before it starts. You always get a fixed quote before work begins.
Who actually writes the code?
Founder-led — no offshoring, no junior handoffs. You work directly with the person designing and building your product. For larger engagements we bring in vetted senior partners we've worked with for years, never blind hires.
Do you work with clients outside South Africa?
Yes — across Africa, the UK and the US. We're GMT+2, with comfortable overlap for EU mornings and US afternoons. Async-first by default, with weekly live syncs.
Can you make AI useful for my business?
Probably — but only where it earns its place. We don't bolt chatbots onto homepages. Real wins look like automating manual back-office work, surfacing intelligence from your data, or agentic workflows that compound over time. We'll tell you honestly if AI isn't the right tool.
Blueprint your build.
Configure what you're making and get an honest indicative range on the spot — the same brackets we publish in our FAQ. A fixed quote follows a 30-minute call.
- Build
- Web platform
- Scope
- Standard build
- Extras
- —
- Typical timeline
- 6–10 weeks
Indicative ZAR range, matching our published pricing. Every engagement gets a fixed quote after a 30-minute call. Not sure what you need? Start with the R15,000 audit.
Send this blueprintLet's find the leak.
Tell us what you're trying to fix. You will hear back within one working day from the person who would actually do the work, including an honest answer if we are not the right fit.
hello@agulhascode.co.za +27 79 866 1382